You are on page 1of 34

Statistics and Probability in High School

Statistics and Probability in High School


Carmen Batanero Statistics and
Probability in High
Universidad de Granada, Spain

and

School
Manfred Borovcnik
University of Klagenfurt, Austria

Statistics and probability are fascinating fields, tightly interwoven with the context of
the problems which have to be modelled. The authors demonstrate how investigations
and experiments provide promising teaching strategies to help high-school students
acquire statistical and probabilistic literacy.
Carmen Batanero and
In the first chapter the authors put into practice the following educational principles, Manfred Borovcnik
reflecting their views of how these subjects should be taught: a focus on the most relevant
ideas and postpone extensions to later stages; illustrating the complementary/dual nature
of statistical and probabilistic reasoning; utilising the potential of technology and show
its limits; and reflecting on the different levels of formalisation to meet the wide variety
of students’ previous knowledge, abilities, and learning types.
The remaining chapters deal with exploratory data analysis, modelling information by
probabilities, exploring and modelling association, and with sampling and inference.
Throughout the book, a modelling view of the concepts guides the presentation.
In each chapter, the development of a cluster of fundamental ideas is centred around a
statistical study or a real-world problem that leads to statistical questions requiring data
in order to be answered. The concepts developed are designed to lead to meaningful
solutions rather than remain abstract entities. For each cluster of ideas, the authors review
the relevant research on misconceptions and synthesise the results of research in order to
support teaching of statistics and probability in high school.
What makes this book unique is its rich source of worked-through tasks and its
focus on the interrelations between teaching and empirical research on understanding Carmen Batanero and Manfred Borovcnik
statistics and probability.

ISBN 978-94-6300-622-4

SensePublishers DIVS

Spine
12.522 mm
  Statistics and Probability in High School  
 
 
 
  Statistics and Probability in High School  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Carmen Batanero  
Universidad  de  Granada,  Spain  
 
and  
 
Manfred Borovcnik  
University  of  Klagenfurt,  Austria  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
  A  C.I.P.  record  for  this  book  is  available  from  the  Library  of  Congress.  
 
 
 
 
 
ISBN:  978-­94-­6300-­622-­4  (paperback)  
ISBN:  978-­94-­6300-­623-­1  (hardback)  
ISBN:  978-­94-­6300-­624-­8  (e-­book)  
 
 
 
 
 
Published  by:  Sense  Publishers,  
P.O.  Box  21858,  
3001  AW  Rotterdam,    
The  Netherlands  
https://www.sensepublishers.com/  
 
 
 
 
 
 
 
All  chapters  in  this  book  have  undergone  peer  review.  
 
 
 
 
 
 
 
Printed  on  acid-­free  paper  
 
 
 
 
 
 
 
All  Rights  Reserved  ©  2016  Sense  Publishers  
 
No  part  of  this  work  may  be  reproduced,  stored  in  a  retrieval  system,  or  transmitted  
in  any  form  or  by  any  means,  electronic,  mechanical,  photocopying,  microfilming,  
recording   or   otherwise,   without   written   permission   from   the   Publisher,   with   the  
exception  of  any  material  supplied  specifically  for  the  purpose  of  being  entered  and  
executed  on  a  computer  system,  for  exclusive  use  by  the  purchaser  of  the  work.  
TABLE OF CONTENTS

Preface     ix  

1.  Educational  Principles  for  Statistics  and  Probability   1  


  1.1.    Introduction   1  
  1.2.    Fundamental  Ideas  in  Statistics  and  Probability   2  
         1.2.1.   Exploratory  Data  Analysis  (Chapter  2)   3  
         1.2.2.   Modelling  Information  by  Probabilities  (Chapter  3)   4  
         1.2.3.   Exploring  and  Modelling  Association  (Chapter  4)   5  
         1.2.4.   Sampling  and  Inference  (Chapter  5)   6  
  1.3.     Complementary  Views  of  Statistics  and  Mathematics   7  
  1.4.     The  Role  of  Technology   10  
  1.5.     Adapting  the  Levels  of  Formalisation  to  the  Diversity  of  Students   12  
  1.6.     Statistical  and  Probabilistic  Literacy   12  
         1.6.1.   Statistical  Literacy   12  
         1.6.2.   Statistical  Literacy  Components   13  
         1.6.3.   Actions  and  Resources  Directed  to  Increase  Statistical  Literacy   14  
  1.7.     Statistical  and  Probabilistic  Thinking   15  
          1.7.1.   Statistical  Thinking  
  16  
          1.7.2.   The  Statistical  Investigation  Cycle    
  16  
          1.7.3.   Fundamental  Types  of  Statistical  Thinking  
  18  
          1.7.4.   Components  of  Probabilistic  Thinking    
  20  
  1.8.     Making  Sense  of  Statistics  and  Probability   21  
  1.9.     Statistical  Investigations  and  Experiments   22  
  1.10.  Final  Thoughts   22  

2.  Exploratory  Data  Analysis   25


  2.1.    Introduction   25  
  2.2.    A  Teaching  Situation  to  Introduce  Elementary  Statistical  Concepts    
and  Procedures   26  
          2.2.1.   Starting  Questions  
  26  
          2.2.2.   Exploring  Qualitative  Variables  
  27  
          2.2.3.   Exploring  Numerical  Variables  
  28  
          2.2.4.   Comparing  Groups  
  34  
  2.3.   Additional  Activities   38  
          2.3.1.   Exploring  Continuous  Variables  
  38  
          2.3.2.   Exploring  Bivariate  Relationships  
  45  
  2.4.   Synthesis  of  Learning  Goals   46  
          2.4.1.   Distribution  and  Different  Types  of  Frequencies  
  47  
          2.4.2.   Simple  Univariate  Graphs  
  48  
          2.4.3.   Simple  Summary  Statistics  
  49  
          2.4.4.   Spirit  of  Exploratory  Data  Analysis  (EDA)  
  51  
          2.4.5.   Basic  Strategies  in  Data  Exploration  
  52  
 
  v  
TABLE  OF  CONTENTS  

  2.5.    Students’  Reasoning  and  Potential  Difficulties   54  


          2.5.1.   Graphical  Competencies  and  Communication  Skills  
  54  
          2.5.2.   Errors  in  Producing  Graphs  
  56  
          2.5.3.   Understanding  Measures  of  Central  Tendency  or  Centre  
  57  
          2.5.4.   Understanding  Spread  
  60  
          2.5.5.   Understanding  Order  Statistics  
  61  
  2.6.   Additional  Resources     62  
          2.6.1.   Journals  and  Books  
  62  
          2.6.2.   Data  Sets  
  63  
          2.6.3.   Internet  Resources  
  63  
          Appendix:  Data  
  65  

3.  Modelling  Information  by  Probabilities 67  


  3.1.    Introduction   67  
  3.2.    Teaching  Situations  to  Characterise  Probability   69  
          3.2.1.   Frequentist  Probability:  Investigating  Coin  Tossing  
  69  
          3.2.2.   Subjectivist  Probability:  The  Insurance  Contract    
  72  
          3.2.3.   Laplace  (A  Priori)  Probability:  Calibrating  Weights  of  Evidence   73  
 

  3.3.    Teaching  Situations  Introducing  Conditional  Probability   74  


          3.3.1.   Conditional  Probability  and  Circumstantial  Evidence  
  75  
          3.3.2.   Conditional  Probability  and  Compound  Probability  
  78  
  3.4.    Additional  Teaching  Activities   79  
          3.4.1.   Random  Variables  
  79  
          3.4.2.   Additivity  of  Expected  Value  and  Variance  for  Repeated    
 

Experiments   81  
          3.4.3.   Distribution  Models  for  Standard  Situations  
  83  
          3.4.4.   Central  Theorems  
  88  
  3.5.    Synthesis  of  Learning  Goals   92  
          3.5.1.   Concepts  to  Model  and  Investigate  Uncertain  Situations  
  92  
          3.5.2.   Different  Connotations  of  Probability  
  93  
          3.5.3.   Circumstantial  Evidence  and  Bayes’  Formula    
  97  
          3.5.4.   Random  Variables  and  Expectation  
  99  
          3.5.5.   Standard  Models  of  Distributions  
  100  
          3.5.6. Law  of  Large  Numbers  and  Central  Limit  Theorem  
  101  
  3.6.    Students’  Reasoning  and  Potential  Difficulties   104  
          3.6.1.   Misconceptions  and  Heuristics  (Strategies)  in  Probability    
 

Situations     104  
          3.6.2.   Independence  and  Conditional  Probability  
  107  
          3.6.3.   Taking  into  Account  Students’  Reasoning  to  Improve  Teaching   108  
 

  3.7.    Additional  Resources  and  Ideas   109  


          3.7.1.   Investigating  Randomness:  Generating  Coin  Tossing  from    
 

Memory     109  
          3.7.2.   Odds  and  Bayes’  Formula  –  Revising  Weights  of  Evidence  
  109  
          3.7.3.   Mediating  Tools  to  Support  Teaching  
  110  
 

vi  
TABLE  OF  CONTENTS  

4.  Exploring  and  Modelling  Association   117  


  4.1.    Introduction   117  
  4.2.    A  Teaching  Situation  to  Explore  Contingency  Tables   119  
          4.2.1.   Exploring  Association  in  2×2  Contingency  Tables  
  119  
          4.2.2.   Different  Distributions  in  2×2  Contingency  Tables  
  121  
          4.2.3.   Simple  Methods  to  Evaluate  Association  in  2×2  Tables  
  123  
          4.2.4.   Expected  Frequencies  for  Independent  Variables  
  124  
  4.3.    Life  Expectancy:  A  Teaching  Situation  to  Explore  Correlation  and  
Regression   126  
          4.3.1.   Exploring  and  Explaining  Correlation  
  127  
          4.3.2.   Fitting  Models  to  Bivariate  Data  
  134  
  4.4.    Additional  Activities   140  
  4.5.    Synthesis  of  Learning  Goals   142  
          4.5.1.   Contingency  Tables  
  143  
          4.5.2.   Correlation  and  Regression  
  144  
  4.6.    Some  Potential  Difficulties   148  
          4.6.1.   Intuitive  Strategies  in  Contingency  Tables  
  148  
          4.6.2.   Linear  Regression  and  Correlation  
  150  
          4.6.3.   Misconceptions  Related  to  both  Association  and  Correlation  
  153  
  4.7.    Additional  Resources  and  Ideas   157  
          4.7.1.   Measures  of  Association  in  Contingency  Tables  
  157  
          4.7.2.   Introduction  to  Multivariate  Visualisation  
  158  

5.  Sampling  and  Inference 163  


  5.1.   Introduction   163  
  5.2.   A  Teaching  Situation:  The  Tea-­Tasting  Experiment   164  
          5.2.1.   The  Lady  Tasting  Tea  
  165  
          5.2.2.   Using  Experimental  Data  to  Test  a  Hypothesis  
  165  
          5.2.3.   Different  Approaches  to  Compute  the  p-­value  
  169  
          5.2.4.   Sampling  Distribution  for  the  Proportion  and  the  Effect  of    
 

Sample  Size   171  


          5.2.5.   Estimating  the  Population  Proportion  by  a  Confidence  Interval  
  172  
  5.3.    Additional  Activities   173  
          5.3.1.   Exploring  the  Central  Limit  Theorem  
  173  
          5.3.2.   Inference  for  Proportions  
  174  
          5.3.3.   Inference  for  a  Mean  
  176  
          5.3.4.   Statistical  Tests  as  Decision  Rules  
  177  
  5.4.    Synthesis  of  Learning  Goals   181  
          5.4.1.   Basic  Inferential  Concepts  and  Procedures  
  181  
          5.4.2.   Additional  Advanced  Learning  Goals  
  184  
  5.5.    Some  Potential  Difficulties   184  
          5.5.1.   Understanding  Random  Sampling  and  Sampling  Distributions  
  185  
          5.5.2.   Understanding  Statistical  Tests  
  187  
          5.5.3.   Understanding  Confidence  Intervals  
  189  

vii  
TABLE  OF  CONTENTS  

  5.6.    Additional  Resources  and  Ideas   191  


          5.6.1.   Developing  Informal  Ideas  of  Inference    
  191  
          5.6.2.   Resampling  Methods  
  191  
          5.6.3.   Updating  a  Prior  Distribution  for  a  Proportion  
  193  

References   197  

Author  Index   209  

Subject  Index   213  


 

viii  
PREFACE

Research   in   statistics   and   probability   education   has   produced   a   variety   of   results  


that  would  be  useful  for  both  secondary  and  high-­school  mathematics  teachers  and  
the   educators   of   these   teachers.   Although   there   are   many   good   textbooks   in  
different  countries  that  describe  statistical  ideas  with  a  formalisation  level  adequate  
for   students,   usually   these   textbooks   are   written   in   a   sequential   way   so   that   the  
different   concepts   and   procedures   are   introduced   in   turn,   with   insufficient  
connections  between  them  and  limited  attention  to  students’  underlying  intuitions.  
  There   are,   of   course,   excellent   exceptions   such   as   the   books   produced   by   the  
Schools   Council   (1980)   in   Statistical   Education   Project   in   the   1980’s;;   yet,   even,  
these   textbooks   do   not   include   a   detailed   summary   of   research   related   to   the  
teaching  of  the  concepts,  which  started  to  get  shape  only  after  the  first  International  
Conference  on  Teaching  Statistics  in  Sheffield  in  1982.  
  In  the  later  stages  of  our  careers  and,  after  collaborating  and  corresponding  for  
many  years  in  different  projects,  we  decided  to  write  a  book  directed  to  reinforce  
the  mathematical  and  didactical  knowledge  of  high-­school  teachers  in  statistics  and  
probability.   At   the   same   time,   we   wish   to   offer   examples   of   potential   activities  
useful   to   introduce   the   main   statistics   and   probability   concepts   and   enhance   the  
underlying  ideas  at  this  school  level.  
  Consequently,  in  this  book  we  provide  examples  of  teaching  situations,  while  at  
the   same   time   we   review   research   on   adolescents’   stochastical1   reasoning   and  
literacy,   with   the   aim   to   provide   recommendations   and   orientations   for   teaching  
these  topics  within  high-­school  mathematics.  The  expression  “high  school”  relates  
to   different   educational   levels   depending   on   the   country;;   in   this   book,   we   will  
consider  students  from  ages  14  to  18  (grades  9–12  in  the  United  States  of  America  
curriculum).  The  book  is  organised  in  five  chapters:  
  In   the   first   chapter,   we   present   some   principles   we   use   to   select   the   content  
analysed   in   the   book   and   the   approach   to   teach   this   content.   These   principles  
emerge  from:    
a.    Our  own  teaching  and  research  experience;;    
b.    An   analysis   of   stochastic   high-­school   curricula   in   several   countries   (e.g.,  
ACARA,  2010;;  NCTM,  2000;;  CCSSI,  2010,  MEC,  2007);;    
c.    The   synthesis   of   available   research   (as   summarised,   for   example,   in   Biehler,  
Ben-­Zvi,   Bakker,   &   Makar,   2013;;   Chernoff   and   Sriraman,   2014;;   Garfield   &  
Ben-­Zvi,   2008;;   Jones,   2005;;   Jones,   Langrall,   &   Money,   2007;;   Shaughnessy,  
1992,  2007;;  Shaughnessy,  Garfield,  &  Greer,  1996);;    

                                                                                                                         
1
    In   some   countries   the   term   stochastics   is   used   to   highlight   the   mutual   dependence   between  
probabilistic   and   statistical   knowledge   and   reasoning.   Throughout   the   book   we   occasionally   use  
stochastics   for   statistics   and   probability   to   express   our   view   that   these   fields   are   tightly  
interconnected  and  should  be  taught  together.  

  ix  
PREFACE  

d.    Our   own   conceptions   of   statistical   and   probabilistic   literacy,   thinking,   and  


reasoning;;  and    
e.    Our   extensive   experience   with   strategies   that   may   help   support   student  
development  in  stochastic  literacy,  thinking,  and  reasoning.  
  The   first   chapter   sets   out   key   educational   principles.   Each   of   the   following  
chapters   (Chapters   2–5)   has   a   focus   on   a   group   of   related   fundamental   stochastic  
ideas,   while   taking   into   account   that   high-­school   stochastics   should   be   built   on  
basic   ideas   that   students   have   encountered   at   primary   and   middle-­school   levels.  
These   chapters   are   organized   according   to   a   common   structure,   including   an  
introduction,   with   a   short   analysis   of   the   main   stochastic   ideas   in   the   particular  
topic   and   its   place   in   the   curriculum;;   some   initial   and   more   advanced   specific  
examples   that   may   serve   to   involve   learners   actively   as   they   progress   in   their  
development   of   the   concepts,   a   summary   of   what   is   known   about   difficulties  
students  encounter  with  the  related  concepts,  a  synthesis  of  the  main  learning  goals  
in   the   chapter,   and   finally,   some   additional   resources   that   may   help   teachers   and  
students.  When  possible,  we  make  connections  between  the  different  chapters  and  
include   some   historical   notes   that   shed   light   on   ways   of   thinking   about   the  
concepts.    
  We  have  tried  to  give  a  balanced  view  on  probability  and  statistics,  with  a  focus  
on  the  interrelated  nature  of  the  concepts,  integrating  probabilistic  ideas  at  a  level  
suitable   for   high   school   teaching,   including   the   step   from   descriptive   statistics   to  
statistical   inference.   Where   ever   we   could   do   it,   we   have   also   tried   to   integrate  
mathematical   concepts   and   contexts   so   that   the   mathematics   developed   becomes  
meaningful   for   the   learners.   May   our   exposition   contribute   to   an   increase   in  
statistical  and  probabilistic  literacy  in  our  societies.    
  We   hope   the   book   will   be   both   useful   for   practising   teachers,   as   well   as   for  
researchers   in   statistics   education   and   practitioners   in   teacher   educators   (teacher  
trainers).   The   different   chapters   contain   original   materials,   but   build   upon   our  
extended  set  of  publications,  part  of  which  is  listed  in  the  references.    
  We  thank  our  colleagues  and  students  who  have  commented  several  drafts  of  the  
chapters.  Among  them  we  want  to  name  especially  two  who  accompanied  us  in  our  
research  work  now  for  decades:  Juan  D.  Godino  and,  particularly,  Ramesh  Kapadia  
who  was  also  helpful  for  improving  the  English.  Finally,  we  would  like  to  express  
our   deepest   gratitude   to   our   families   and   friends   for   their   encouragement   and  
support  over  the  years  when  we  were  writing  the  book.  
 
May,  2016  
Carmen  Batanero  and  Manfred  Borovcnik  

x  
CHAPTER  1  

EDUCATIONAL PRINCIPLES FOR


STATISTICS AND PROBABILITY
 

In   this   chapter,   we   describe   those   principles,   which   reflect   our   view   of   how   stochastics  
should   be   taught   at   high-­school   level.   Firstly,   we   suggest   the   need   to   focus   on   the   most  
relevant  ideas  for  the  education  of  students.  Secondly,  we  analyse  the  complementary  nature  
of  statistical  and  mathematical  reasoning  on  the  one  side  and  of  statistical  and  probabilistic  
reasoning   on   the   other   side.   We   then   examine   the   potential   and   limits   of   technology   in  
statistics  education  and  reflect  on  the  different  levels  of  formalisation  that  may  be  helpful  to  
meet   the   wide   variety   of   students’   previous   knowledge,   abilities,   and   learning   types.  
Moreover,  we  analyse  the  ideas  of  probabilistic  and  statistical  literacy,  reasoning  and  sense  
making.   Finally,   we   demonstrate   how   investigations   and   experiments   provide   promising  
teaching  strategies  to  help  high-­school  students  to  develop  these  capabilities.    

1.1.  INTRODUCTION  

We   are   surrounded   by   uncertainty   that   affects   our   lives   in   personal,   biological,  


social,   economic,   and   political   settings.   This   fact   suggests   that   we   need   to  
understand  random  phenomena  to  make  adequate  decisions  when  confronted  with  
uncertainty.   We   meet   arguments   based   on   data   in   everyday   life;;   to   critically  
evaluate   these   arguments,   we   need   to   understand   the   way   in   which   the   data   is  
produced  and  how  the  conclusions  are  obtained.  
  Three   widely   accepted   reasons   for   including   stochastics   at   the   school   level   are  
the  essential  role  of  statistics  and  probability  in  critical  reasoning,  its  instrumental  
role  in  other  disciplines,  and  its  key  role  for  planning  and  decision  making  in  many  
professions   (Franklin   et   al.,   2007;;   Gal,   2002,   2005;;   Wild   &   Pfannkuch,   1999).  
These   reasons   have   been   recognised   for   some   time   and,   hence,   the   teaching   of  
statistics  at  secondary  school  has  a  history  of  about  30  years  in  many  countries,  for  
example,   in   Australia,   Austria,   France,   Israel,   Germany,   Spain,   and   in   the   US.   A  
recent  innovation  is  the  extension  of  statistics  teaching  to  lower  grades  so  that  it  is  
now   included   throughout   the   curriculum   starting   from   the   first   year   of   primary  
school  to  graduate  courses  and  postgraduate  training  at  universities.  
  Statistical   ideas   originate   from   the   natural   sciences   and   demography   and   have  
transformed   many   fields   of   human   activity   over   the   past   three   centuries  
(Gigerenzer   et   al.,   1989;;   Hacking,   1990).   Today,   statistics   is   pervasive;;   there   is  
scarcely  a  political,  scientific,  or  social  issue  without  reference  to  statistical  results.  
People   encounter   statistical   information   while   shopping,   watching   TV,   reading   a  
newspaper,   or   surfing   on   the   Internet.   Furthermore,   national   statistics   offices   and  
international   agencies   such   as   the   United   Nations   (UN)   or   the   World   Health  
Organisation  (WHO)  make  their  statistical  studies  available  on  the  Internet.    
  1  
CHAPTER  1  

  Statistical   methods   are   important   not   only   in   various   disciplines   in   science   but  
also   for   government   and   business   systems,   e.g.,   health   records   or   retirement  
pension   plans.   It   is   essential   to   understand   statistics   and   probability   to   critically  
evaluate   how   statistics   are   generated   and   to   justify   decisions,   be   they   societal   or  
personal  (Hall,  2011).  Accordingly,  statistics  educators,  educational  authorities,  as  
well   as   statistical   offices   and   societies   call   for   a   statistically   literate   society   and  
support  projects  that  help  children  and  adults  to  acquire  the  competencies  needed  
in  the  era  of  data  information.    
  The   aim   of   this   chapter   is   to   clarify   what   is   meant   by   statistical   literacy,  
statistical   thinking,   statistical   reasoning,   and   sense   making.   These   aspects   of  
learning   statistics   have   been   widely   discussed   in   the   Statistical   Reasoning,  
Thinking,   and   Literacy   (SRTL)   Research   Forum,   a   series   of   conferences  
(srtl.fos.auckland.ac.nz/)  starting  in  1999  as  well  as  in  Ben-­Zvi  and  Garfield  (2004)  
and   Garfield   and   Ben-­Zvi   (2008).   We   also   suggest   that   students   can   acquire  
relevant  competencies  through  statistical  projects  and  investigations.  
  Teaching  statistics  and  probability  at  high-­school  level  is  often  embedded  within  
mathematics.   However,   due   to   its   peculiarities,   statistics   and   probability   require  
special  attention  on  the  part  of  teachers  and  curriculum  designers  in  relation  to  the  
selection  of  content  and  the  best  way  to  make  the  statistical  ideas  accessible  to  the  
students.  The  goal  of  Chapter  1  is  to  present  our  overall  perspective  on  the  teaching  
of  statistics  and  probability.  This  perspective  is  made  more  explicit  in  Chapters  2  to  
5  that  deal  with  the  teaching  of  the  main  ideas  of  this  subject  at  high-­school  level.  

1.2.  FUNDAMENTAL  IDEAS  IN  STATISTICS  AND  PROBABILITY  

Given  that  the  time  available  for  teaching  is  limited,  it  is  important  to  select  the  key  
concepts   that   should   be   taught   carefully.   Several   authors   (e.g.,   Borovcnik   &  
Kapadia  2014a;;  Borovcnik  &  Peard,  1996;;  Burrill  &  Biehler,  2011;;  Heitele,  1975)  
have  investigated  the  history  of  statistics,  the  different  epistemological  approaches  
to   probability   and   statistics,   and   the   curricular   recommendations   in   different  
countries,   as   well   as   the   educational   research.   Based   on   their   studies,   they   have  
proposed   various   lists   of   “fundamental”   ideas.   These   ideas   can   be   taught   at  
different   levels   of   formalisation   depending   on   students’   ages   and   previous  
knowledge  (see,  e.g.,  Borovcnik  &  Kapadia,  2011).    
  Our   suggestions   to   teach   the   statistics   content   (Chapters   2   to   5)   are   organised  
around  four  main  clusters  of  fundamental  ideas.  We  consider  these  clusters  as  key  
foci  around  which  activities  can  be  organised  by  teachers  in  order  to  help  students  
acquire  the  related  key  concepts.  Each  chapter  analyses  the  essential  content  of  one  
cluster   using   paradigmatic   problems   or   situations   that   can   be   put   forward   to   the  
students.   When   appropriate,   we   make   connections   across   the   content   in   the   other  
chapters.   In   these   chapters,   we   also   inform   teachers   about   the   learning   goals  
implicit  in  the  activities,  point  out  potential  difficulties  encountered  by  learners  as  
described  by  research,  and  suggest  promising  teaching  resources  and  situations  that  
embed  the  ideas  within  instruction.  A  summary  of  the  fundamental  ideas  included  
in  each  of  these  chapters  follows.  

2  
EDUCATIONAL  PRINCIPLES  FOR  STATISTICS  AND  PROBABILITY  

1.2.1.   Exploratory  Data  Analysis  (Chapter  2)  


Basically,  statistics  deals  with  collecting  and  analysing  data,  and  making  decisions  
based  on  data.  The  starting  point  in  a  statistical  study  is  a  real-­world  problem  that  
leads   to   some   statistical   questions   requiring   data   in   order   to   be   answered.   To  
address   the   questions,   such   data   may   already   exist;;   yet,   often   new   data   must   be  
produced   to   provide   sufficiently   valid   and   reliable   information   to   make   a  
judgement  or  decision.  Unlike  mathematics,  answers  to  statistical  questions  always  
involve   an   element   of   uncertainty.   Furthermore,   in   contrast   to   mathematics,   in  
statistics  the  context  of  the  data  is  a  critical  component  in  the  investigations.    
  For  example,  when  studying  the  volume  of  a  cylinder  in  mathematics,  the  same  
formula   always   applies   no   matter   whether   the   cylinder   is   a   juice   can   or   part   of   a  
building.   In   statistics,   however,   the   type   of   data   and   the   context   are   essential   for  
choosing   the   appropriate   method   of   analysis   and   for   interpreting   the   results.   The  
challenge   of   statistical   problem   solving   is   the   need   to   interpret   the   statistical  
concepts   within   a   given   context   and   choose   the   most   suitable   method   from   a  
variety   of   possible   methods   that   may   be   applied   for   the   problem.   Although   the  
computations   in   statistical   procedures   can   often   be   completely   outsourced   to  
software,  the  decision  about  which  procedures  and  techniques  should  be  used  and  
the   interpretation   of   the   results   within   the   context   of   the   data   remain   a   big  
challenge  when  teaching  statistics.    
  Today,  there  is  a  large  amount  of  data  accessible  on  the  Internet  on  almost  every  
topic   that   may   interest   students.   This   helps   to   facilitate   working   with   real   data,  
which  can  increase  students’  levels  of  motivation  (see,  e.g.,  Hall,  2011;;  Ridgway,  
2015).  Real  data  sets  also  help  students  investigate  issues  that  are  rarely  mentioned  
in   traditional   problems   in   textbooks.   For   example,   students   can   explore   different  
ways  of  collecting  data,  design  their  own  questionnaires  or  experiments,  and  gain  a  
better  understanding  of  different  data  types.  When  doing  their  own  investigations,  
students   can   encounter   additional   problems   in   managing   missing   or   incomplete  
data,   come   across   data   that   are   atypical   (or   even   wrong   due   to   a   failure   in   the  
collection   process).   They   have   to   refine   the   data   set   accordingly,   or   assess   the  
reliability  and  validity  of  the  investigated  data.  

Representations  of  data  play  a  major  role  in  statistics  as  a  variety  of  graphs  can  be  
used   to   display   and   extract   information   that   may   be   hidden   in   the   raw   data.   The  
process  of  changing  the  representation  of  data  in  order  to  find  further  information  
relevant  to  the  initial  problem  is  called  transnumeration  and  is  considered  to  be  an  
important  process  in  statistical  reasoning  (Wild  &  Pfannkuch,  1999).    

Variation  and  distribution  are  two  complementary  concepts  that  play  a  vital  role  in  
statistics.   Although   variables   and   variation   also   appear   in   many   areas   of  
mathematics,   mathematics   focuses   on   functional   (deterministic)   variation   while  
statistics   deals   with   random   variation.   Hence,   a   goal   of   statistics   education   is   to  
enable  students  to  reason  about  data  in  context  under  conditions  of  uncertainty,  and  
to  discriminate  between  statistical  reasoning  and  mathematical  reasoning.  Wild  and  

3  
CHAPTER  1  

Pfannkuch   (1999)   suggest   that   the   perception   of   random   variation   is   an   essential  


component   of   statistical   thinking.   Moreover,   statistics   provides   methods   to  
identify,  quantify,  explain,  control,  and  reduce  random  variation.  

Distribution  is  a  term  that  is  specific  to  statistics  and  probability;;  it  is  a  collection  
of   properties   of   a   data   set   as   a   whole,   not   of   a   particular   value   in   the   data   set.   A  
distribution  consists  of  all  the  different  values  in  the  data  including  the  frequencies  
(or  probabilities)  associated  with  each  possible  value.  Variation  and  distribution  are  
linked  to  other  fundamental  statistical  ideas  such  as  centre  (as  modelled  by  mean,  
median,   or   mode),   spread   (as   modelled   by   standard   deviation   or   variance),   and  
shape   (for   example,   bi-­modal,   uniform,   symmetric,   or   L-­shaped).   Measures   of  
centre   summarise   the   information   about   a   distribution   while   measures   of   spread  
summarise   the   variability   within   the   data.   Each   value   of   a   variable   shows   some  
deviation   from   the   centre.   In   the   context   of   measuring   an   unknown   quantity  
(signal),   this   deviation   may   be   interpreted   as   an   error   in   measurement   (noise).  
Metaphorically,  the  distribution  embodies  an  overall  “model”  for  potential  errors  or  
noise,  while  the  centre  can  be  seen  as  the  signal  (Konold  &  Pollatsek,  2002).  

1.2.2.   Modelling  Information  by  Probabilities  (Chapter  3)  


Descriptive  statistics,  probability  theory,  and  statistical  inference  complement  each  
other   and   view   the   same   information   from   different   angles.   Descriptive   statistics  
investigates   the   information   from   one   sample   (data   set)   and   summarises   it   using  
single   numbers   and   graphical   displays.   Statistical   inference   goes   beyond   the  
present   data   set   and   tries   to   generalise   the   findings,   that   is,   to   transfer   them   to   a  
wider   population   to   which   the   data   set   belongs.   Probability   takes   the   role   of   a  
mediator  as  it  supplies  a  justification  for  a  generalisation  beyond  the  initial  data.    
  The  information  extracted  from  the  data  by  using  descriptive  statistics  can  only  
be  generalised  by  inferential  methods  that  are  established  by  probability  models.  At  
the   same   time   some   probability   concepts   are   more   easily   understood   as   a  
generalisation   of   descriptive   statistics   (e.g.,   the   idea   of   expectation   can   be  
understood   as   a   generalisation   of   the   idea   of   mean).   These   links   have   influenced  
and   shaped   school   curricula.   Yet,   the   study   of   statistics   is   incomplete   if   the  
reference   to   probability   is   missing   in   our   teaching.   The   fundamental   purpose   of  
probability   is   twofold:   to   measure   or   to   evaluate   uncertainty.   The   first   idea  
considers  probability  as  a  property  of  physical  objects  like  length,  while  the  second  
notion   points   towards   a   qualitative   procedure   of   subjectively   attaching   numerical  
values  to  uncertain  phenomena.  In  the  book,  we  take  into  account  these  two  ideas  
and  the  different  meanings  of  probability.  
  In  recent  years,  there  has  been  a  shift  in  the  way  probability  is  taught  at  school  
level,   from   a   classical   Laplacean   (or   “axiomatic”)   approach   (common   until   the  
1980s)   towards   a   frequentist   conception   of   probability,   that   is,   an   experimental  
approach   where   probabilities   are   estimated   from   long-­range   relative   frequencies  
(Batanero,  Chernoff,  Engel,  Lee,  &  Sánchez,  2016).  Simulations  and  experiments  
are   used   to   support   students   in   understanding   the   Law   of   Large   Numbers   and  

4  
EDUCATIONAL  PRINCIPLES  FOR  STATISTICS  AND  PROBABILITY  

grasping  the  sophisticated  interaction  between  the  notion  of  relative  frequency  and  
the   frequentist   conception   of   probability.   The   subjectivist   view   of   probability,  
which   is   widely   used   in   applied   statistics,   has   been   developed   hand   in   hand   with  
the  frequentist  view  so  that  the  two  complement  each  other  (Hacking,  1975).  Their  
interplay   is   relevant,   especially   for   conditional   probability.   Bearing   this   in   mind,  
we  suggest  a  combination  of  both  approaches  in  the  teaching  of  probability.  
  The   fundamental   ideas   of   variable   and   distribution   still   apply   for   probabilistic  
modelling.   However,   probability   distributions   deal   with   the   potential   data   of   a  
“random  experiment”,  which  models  how  data  will  be  generated,  rather  than  with  
actual   data   that   have   been   collected.   A   helpful   idea   is   to   think   of   repeated  
experiments   that   supply   us   with   idealised   frequencies.   This   metaphor   helps   us   to  
transfer   many   concepts   from   descriptive   statistics   to   probability   and   delivers   a  
more  concrete  picture   of  what  probability  means.1  A  variety  of  representations  of  
descriptive  statistics  can  be  transferred  to  probabilities.  Furthermore,  tree  diagrams  
may  be  used  to  simplify  the  discussion  of  combined  random  experiments  and  the  
calculus  of  probabilities.  All  of  these  approaches  are  presented  in  Chapter  3.  

1.2.3.   Exploring  and  Modelling  Association  (Chapter  4)  


In   a   statistical   study,   we   often   are   interested   in   checking   whether   two   (or   more)  
variables   are   interrelated   and   whether   some   type   of   function   may   describe   their  
interaction.   Functions   occupy   a   central   place   in   mathematics   and   are   used   to  
describe   connections   between   variables   in   many   fields   (such   as   economy   or  
physics)   in   situations   that   have   the   following   features.   First,   the   function   may   be  
determined  by  general  laws.  Second,  in  practical  experiments  these  functions  may  
only   be   slightly   blurred   by   small   measurement   errors   so   that   the   underlying  
functional   relationship   is   still   visible   from   graphs   of   the   data.   In   physics,   for  
example,   the   value   of   the   independent   variable   determines  a  specific  value  of  the  
dependent  variable.  
  Independent   (explanatory)   and   dependent   (response)   variables   occur   in  
mathematics   and   statistics   but   the   link   between   them   is   not   as   strong   in   the  
statistical   context   as   it   is   in   mathematics.   Data   on   variables   such   as   heights   of  
fathers  and  sons  might  show  a  similar  tendency  but  its  pattern  is  less  clear  than  in  
data   on   variables   that   follows   a   relationship   in   physics;;   yet,   a   description   of   the  
interrelationship   might   be   useful.   When   we   study   the   relation   between   smoking  
and   the   occurrence   of   lung   cancer,   we   deal   with   qualitative   variables.   Moreover,  
while  the  percentage  of  people  with  lung  cancer  is  greater  among  those  who  have  
previously   been   smokers   –   which   indicates   that   it   is   better   not   to   smoke   –   this  
association  cannot  be  interpreted  directly  in  causal  terms  as  it  could  be  induced  by  
third   factors   (hidden   variables)   that   may   be   operating   in   the   background.   One  
example   is   that   problems   in   handling   stress   can   lead   people   to   smoke   and   make  
them  more  prone  to  lung  cancer.  
                                                                                                                         
1
    In  Chapter  3,  we  also  see  that  in  many  situations  probability  models  are  applied  to  one-­off  decisions.  

5  
CHAPTER  1  

  Three   different   problems   can   be   studied   when   modelling   statistical   relations  


between  variables:2  
1. Are   two   variables   interrelated?   This   leads   to   the   concepts   of   association   and  
correlation.  
2. Is   there   a   mathematical   function   that   may   be   used   to   describe   the   relationship  
and  is  it  possible  to  justify  the  criteria  in  selecting  such  a  function?  This  is  called  
the   regression   problem.   At   high-­school   level,   teaching   is   usually   restricted   to  
linear  functions  to  describe  the  relations  between  the  variables  as  they  are  easier  
to  interpret.  
3. How   well   does   a   specific   function   describe   the   relationship   between   the  
variables?   This   leads   to   the   problem   of   finding   a   suitable   measure   of   the  
goodness   of   fit.   For   general   functions,   the   coefficient   of   determination,   which  
coincides   with   the   square   of   the   correlation   coefficient   for   linear   functions,   is  
used.  
  Regression  becomes  more  complex  if  the  influence  of  several  variables  on  one  
(response)  variable  is  analysed.  Therefore,  the  mathematical  techniques  needed  to  
find   suitable   functions   to   model   multivariate   relationships   between   the   target  
variable   and   several   independent   variables   are   usually   outsourced   to   statistical  
packages.  Yet,  people  who  perform  the  analysis  need  to  understand  the  basic  ideas  
behind   these   methods.   They   also   should   understand   that   data   becomes   more  
reliable  if  a  special  design  to  control  for  third  factors  is  followed.  As  the  topic  is  so  
relevant,   elementary   parts   of   it   are   contained   in   school   curricula   with   just   two  
variables   under   scrutiny,   while   supported   by   technology   both   for   the   required  
calculations   and   for   drawing   specific   graphs.   The   interpretation   of   the   results   can  
be  sophisticated  and  is  often  a  challenge  for  teaching.  

1.2.4.   Sampling  and  Inference  (Chapter  5)  


The  fundamental  idea  of  statistical  inference  is  to  generalise  information  from  data  
sets  to  wider  entities.  It  is  related  to  inductive  logic,  which  deals  with  principles  of  
generalising   empirical   information   (Hacking,   1965;;   Popper,   1959).   Although  
mathematics   derives   true   statements   by   rules   of   ordinary   logic,   in   statistical  
inference  the  generalisations  are  constructed  on  the  basis  of  hypotheses  that  include  
probabilities.  In  spite  of  being  a  very  young  field  of  study,3  statistical  inference  has  
paved  the  way  for  our  evidence-­based  society.    
  Fundamental   ideas   of   inference   are   population   and   sample   and   their  
interrelationships.   We   might   observe   a   process   of   data   generation   for   some   time  
(e.g.,  the  weight  of  fabricated  units  in  a  factory)  or  a  sample  of  data  from  a  subset  
of   a   finite   population   (e.g.,   the   weight   of   some   eight-­year-­old   children   in   a  
                                                                                                                         
2
    Chapter   4   is   restricted   to   the   descriptive   study   of   the   topic   although   all   of   these   problems   can   be  
generalised  with  inferential  methods.  
3
    Apart   from   rudimentary   earlier   traces,   the   methods   were   developed   between   1920   and   1950.   The  
successful  axiomatisation  of  probability  by  Kolmogorov  in  1933  increased  the  prestige  of  statistical  
methods,  which  are  today  applied  in  every  area  of  human  activity.    

6  
EDUCATIONAL  PRINCIPLES  FOR  STATISTICS  AND  PROBABILITY  

country).  In  both  examples  we  might  be  interested  in  a  number  that  describes  the  
average   weight   of   the   population.   In   the   first   case   the   focus   is   on   all   future   units  
produced,   whereas   in   the   second,   the   focus   is   all   current   eight-­year-­olds   in   the  
country.    
  An   important   idea   is   that   of   sample   representativeness.   If   there   are   no   biasing  
factors   in   how   elements   are   selected,   then   the   average   weight   of   the   sample   data  
should   be   a   reliable   estimate   of   the   average   (expected)   weight   of   the   population.  
The   statistical   way   to   “guarantee”   representativeness   is   to   control   the   sampling  
process.   More   precisely,   random   sampling   is   a   suitable   method   to   obtain   a  
representative  sample,  and  it  is  possible  to  calculate  the  probability  of  obtaining  a  
biased  sample.  The  techniques  for  generalising  the  information  from  samples  to  the  
whole  population  are  confidence  intervals  and  tests  of  hypotheses.  
  Today,   statistical   inference   has   found   its   way   into   curricula   all   over   the   world  
with   a   variety   of   approaches   that   attempt   to   make   the   methods   and   the   inherent  
notions  more  accessible  to  students.  Specifically  it  is  common  to  use  simulation  to  
facilitate   parts   of   the   computation   and   to   visualise   the   sampling   variability.   More  
recently,  resampling  approaches  have  been  used  to  simplify  the  probability  models  
implicit   in   inference   methods   by   focussing   entirely   on   the   data   set   that   has   to   be  
analysed.  All  of  these  methods,  as  well  as  Bayes’  rule  to  update  information  from  
empirical  data,  are  presented  in  Chapter  5.  
  In  conclusion,  the  most  fundamental  objective  in  probability  and  statistics  is  to  
offer   models   to   understand   and   interpret   real   observations.   A   model   does   not  
completely   represent   reality;;   yet,   it   can   be   used   for   explorations   that   may   lead   to  
relevant   results   or   insights   about   real   data.   A   fundamental   goal   of   teaching  
statistics   is  that  students  understand   the  hypothetical  character  of   probability  and  
statistical   models   and   the   possibility   of   applying   these   models   in   many   different  
contexts.    

1.3.  COMPLEMENTARY  VIEWS  OF  STATISTICS  AND  MATHEMATICS  

Today,   the   teaching   of   statistics   at   university   level   is   often   separated   from  


mathematics.   For   example,   in   countries   like   Spain   or   the   US,   distinct   degrees   or  
graduate   programmes   are   offered   in   the   training   of   mathematicians   and  
statisticians.   In   other   countries   such   as   Austria   or   Germany,   there   are   stronger  
connections  between  mathematics  and  statistics.    
  Research   in   statistics   and   probability   (stochastics)   is   promoted   by   the  
International  Statistical  Institute  and  other  organisations  with  specific  conferences  
(e.g.,   the   World   Statistics   Congress)   and   journals   (Annals   of   the   Institute   of  
Statistical   Mathematics   or   Computational   Statistics).   Statistics   education   research  
has   received   wide   input   from   areas   different   from   mathematics,   for   example,  
psychology,   mathematics   education,   and   general   education   (see   Batanero,  
Chernoff,  Engel,  Lee,  &  Sánchez,  2016;;  Jones,  2005;;  Jones,  Langrall,  &  Mooney,  
2007;;   Shaughnessy,   1992,   2007;;   Shaughnessy,   Garfield,   &   Greer,   1996;;   Vere-­
Jones,   1995,   for   a   detailed   description).   As   discussed   in   Section   1.6.3,   more  

7  
CHAPTER  1  

recently   the   field   of   statistics   education   has   grown   to   become   a   discipline   in   its  
own  right.4  
The   specific   character   of   statistics   and   probability   is   also   reflected   in   the  
philosophical,  ethical,  procedural,  and  even  political  debates  that  are  still  ongoing  
within   these   areas   and   their   applications,   which   does   not   often   happen   in  
mathematics.5  Statistics  and  probability  are  closely  related  to  other  sciences  such  as  
demography,  genetics,  insurance,  astronomy,  and  law,  from  which  many  statistical  
methods   were   developed.   Furthermore,   inferential   statistics   has   formed   the   basis  
for   a   new   scientific   paradigm   in   social   sciences,   medicine,   and   other   disciplines.  
These   disciplines   share   the   process   of   a   scientific   argument   for   generalising  
empirical  results  that  leads  beyond  the  subjectivity  and  the  restrictions  of  a  single  
experimental  situation.    
Subsequently,  we  describe  the  components  of  an  empirical  study,  which  starts  
with  a  contextual  question  (in  Section  1.7.2).  This  question  leads  to  an  appropriate  
design  with  a  corresponding  statistical  question  as  well  as  a  plan  on  how  to  collect  
the   data   needed   to   address   the   question   according   to   the   chosen   design.   This  
careful  planning  is  necessary  in  order  to  obtain  useful  information  about  the  initial  
question   and   to   keep   the   information   free   from   confounding   effects.   The  
exploration  and  analysis  of  the  data  are  followed  by  drawing  some  conclusion  from  
the  data  (Wild  &  Pfannkuch,  1999).  As  we  expose  in  Chapter  2,  a  crucial  final  step  
is  the  interpretation  of  the  results  in  relation  to  the  initial  question  given  the  context  
of   the   problem   (see   also   the   steps   outlined   in   a   statistical   investigation   in   the  
GAISE  project  in  Franklin  et  al.,  2007).    
  The   main   interest   in   applying   statistics   in   a   real-­world   study   concerns   finding,  
describing,   and   confirming   patterns   in   data   that   go   beyond   the   information  
contained   in   the   available   data.   Thus,   statistics   is   often   viewed   as   the   science   of  
data   or   as   a   tool   for   conducting   quantitative   investigations   of   phenomena,   which  
requires   a   well-­planned   collection   of   observations   (Gattuso   &   Ottaviani,   2011;;  
Moore  &  Cobb,  2000).  This  feature  of  statistics  explains  why  it  is  easy  to  establish  
connections  between  statistics  and  other  school  subjects  and  why  it  has  sometimes  
been   argued   that   statistics   should   be   taught   outside   the   mathematics   classroom  
(Pereira-­Mendoza,  1993).  

Statistics   is   part   of   the   mathematics   curricula.   As   argued   by   Usiskin   (2014),  


statistics  involves  a  great  deal  of  mathematical  knowledge.  Fitting  statistics  within  
mathematics   teaching   means   that   no   additional   time   is   needed   for   a   separate  
subject   in   school,   where   time   is   limited.   According   to   Usiskin,   as   well   as   to  
Scheaffer   (2006),   statistics   and   mathematics   may   support   and   complement   one  
another  in  the  school  curriculum.    

                                                                                                                         
4
    For  example,  in  the  US,  there  are  graduate  programmes  in  statistics  education  and  doctoral  degrees  
have  been  awarded  in  statistics  education.  
5
    An  example  is  the  controversy  around  the  use  of  statistical  tests  (Batanero,  2000;;  Borovcnik,  1986a;;  
Hacking,  1965).  

8  
EDUCATIONAL  PRINCIPLES  FOR  STATISTICS  AND  PROBABILITY  

  An  example  described  by  Usiskin  is  the  natural  extension  from  the  straight  line  
that   goes   exactly   through   two   points   to   linear   regression   where   the   interest   is   in  
finding  a  line  that  fits  to  more  than  two  data  points  in  an  optimal  way.  Finding  the  
line   of   best   fit   is   part   of   mathematical   modelling;;   however,   statistical   modelling  
does  not  stop  there.  Lines  are  compared  to  other  functions  that  may  also  model  the  
data   and   the   interpretation   of   the   fitted   function   depends   on   the   context.   Typical  
questions  are:    
  Do   other   functions   fit   better?   What   does   it   mean   if   we   describe   the   relation  
between  the  two  variables  under  scrutiny  by  a  line?  Can  we  predict  the  value  of  the  
dependent  variable  outside  the  range  of  data?    
  According   to   Usiskin,   statistics   education   can   also   benefit   from   the   strong  
movement  towards  modelling  developed  since  the  late  1980s  especially  promoted  
by   the   ICTMA,   the   International   Community   of   Teachers   of   Mathematical  
Modelling   and   Applications,6   and   introduced   only   more   recently   in   statistics  
education   through   the   Exploratory   Data   Analysis   (EDA)   movement.7   In   this  
modelling   approach,   the   motivation   for   a   particular   concept   emerges   from   the  
context;;  the  concepts  are  developed  interactively  by  contextual  considerations  and  
mathematical   principles.   The   teacher   should   find   an   appropriate   real   situation   in  
which   the   new   concept   makes   sense   for   the   students.   Throughout   the   book   we  
present   such   initial   problems   that   help   students   understand   the   related   concepts,  
followed  by  more  complex  situations  when  space  permits.  
  Statistics  at  high  school  may  sometimes  be  a  separate  course  (e.g.,  the  advanced  
placement   course   in   the   US)   or   included   in   other   subjects   and   taught   when   it   is  
needed  to  understand  the  current  topics.  We  agree  with  Usiskin  that  regardless  of  
the  placement,  all  students  should  experience  substantial  school  work  in  statistics.  
In   this   way,   they   become   competent   to   appreciate   and   criticise   empirically-­based  
arguments  around  them  and  empowered  to  make  adequate  and  informed  decisions.  

Probability   versus   statistics.   Some   authors   consider   that   probability   is   more  


strongly   linked   to   mathematics   than   to   statistics.   Throughout   the   book,   we   try   to  
make  clear  that  statistics  and  probability  complement  each  other  and  should  not  be  
completely   separated.   The   frequentist   view   of   probability   serves   to   connect  
probability  with  a  wide  range  of  applications  and  probability  cannot  be  understood  
without   a   connection   to   relative   frequencies.   Moreover,   modelling   probability   as  
the   limit   of   relative   frequencies   provides   a   first   basis   for   introducing   statistical  
inference  from  a  frequentist  approach.    
  The   subjectivist   view   of   probability   extends   the   applications   of   probability   to  
decision   making   in   one-­off   situations   where   the   frequentist   view   does   not   apply.  
Additionally,   Bayes’   rule   establishes   strong   links   between   these   two   perspectives  
on  probability  and  allows  combining  subjective  probabilities  and  statistical  data  to  
                                                                                                                         
6
    The   PPDAC   cycle   of   Wild   &   Pfannkuch   (1999)   in   Section   1.7.2   has   close   connections   to   the  
modelling  cycle.  
7
    Although  EDA  also  started  in  the  1980s,  it  was  then  mainly  a  movement  towards  methods  that  were  
attractive  to  teaching  because  of  their  simplicity.  

9  
CHAPTER  1  

update   prior   probability   judgements   and   make   them   “more   objective”.   Finally,  
some   probability   ideas   are   needed   to   understand   more   informal   approaches   to  
inference   when   we   use   simulations   or   re-­randomisation   to   generate   empirical  
sampling  distributions  (see  Chapter  5).    
  Thus,   when   teaching   probability,   one   has   to   consider   that   probability   is   a  
theoretical  concept  –  a  virtual  concept  according  to  Spiegelhalter  and  Gage  (2014)  
–  and  we  speak  about  probability  by  using  metaphors  such  as  “propensity”,  “degree  
of   belief”,   or   “limit   of   frequencies”,   which   convey   only   parts   of   this   abstract  
concept.  Even  though  the  relationship  between  probability  and  relative  frequencies  
is   fundamental   for   the   comprehension   of   probability   and   statistical   methods,   this  
relationship  is  not  always  well  understood  as  some  students  confuse  frequency  with  
probability.8  Moreover,  the  different  representations  of  measures  of  uncertainty  (as  
absolute   numbers9   versus   probabilities)   may   also   involve   different   levels   of  
difficulty   in   understanding   probability   models.   In   Chapter   3,   we   discuss   other  
difficulties   encountered   by   the   students   when   interpreting   small   probabilities,  
which   usually   occur   in   the   case   of   risks   of   adverse   events   such   as   a   maximum  
credible  accident  of  a  nuclear  power  station  or  dying  from  lung  cancer.    

1.4.  THE  ROLE  OF  TECHNOLOGY  

Technology   has   revolutionised  the  applications  of  statistics   and  likewise   statistics  
education.  With  software  such  as  Fathom  (Finzer,  2007)  or  Tinkerplots  (Konold  &  
Miller,   2011),   specially   designed   to   support   the   learning   of   statistics   and  
probability,   with   a   spreadsheet,   or   even   with   Internet   applets,   data   analysis   is   no  
longer   the   exclusive   domain   of   statisticians   (Biehler,   1997;;   Biehler,   Ben-­Zvi,  
Bakker,   &   Makar,   2013;;   Pratt,   Davies,   &   Connor,   2011).   As   demonstrated  
throughout  Chapters  2  to  5,  software  can  facilitate  computations  and  the  production  
of  graphical  representations  of  data.  Thus,  students  can  use  methods  such  as  fitting  
a   variety   of   models   to   a   scatter   plot   (see   Chapter   4)   that   were   not   accessible   to  
them  a  few  years  ago.    
  With   modern   technology,   students   can   represent   abstract   interrelationships   and  
operations,  interact  with  the  setting,  and  see  the  changes  in  the  representation  or  in  
the   results   when   varying   some   data   or   parameters.10   Technology   provides   the  
possibility   of   dynamic   visualisations   where   the   impact   of   crucial   parameters   on   a  
graph  can  be  traced.  This  technique  may  serve,  for  example,  to  explore  the  waiting  
time  for  the  first  success  in  Bernoulli  experiments  (as  done  in  Chapter  3).    
  Due   to   facility   and   speed   of   computations,   the   size   of   data   sets   is   no   longer   a  
limitation   so   that   it   becomes   easier   to   use   real   data   collected   by   the   students   or  
taken  from  the  Internet  (e.g.,  from  CensusAtSchool,  n.d.,  or  from  several  statistical  
                                                                                                                         
8
    The   expression   “empirical   probability”   is   unfortunate   in   this   regard   as   a   probability   is   always  
theoretical;;  only  the  frequencies  are  empirical.  
9
    In  the  sense  of  Gigerenzer  (1994).  
10
    Another   possibility   not   studied   in   the   book   is   multivariate   dynamic   visualisation   that   can   be  
implemented  with  tools  like  those  available  from  Gapminder  (Rosling,  2009).    

10  
EDUCATIONAL  PRINCIPLES  FOR  STATISTICS  AND  PROBABILITY  

offices).   In   Chapter   2,   we   suggest   that   students   collect   and   analyse   physical  


measures   of   themselves   such   as   height   and   weight,   arm   span,   shoe   size,   etc.  
However,   teachers   should   be   careful   about   issues   such   as   students   being   able   to  
understand   the   data   within   the   context   and   to   connect   the   data   set   to   the  
investigated  problem.  Otherwise  students  may  lose  interest  (Hall,  2011).  
  Another   advantage   of   technology   is   that   it   can   help   build   micro   worlds   where  
students  can  explore  conjectures  and  establish  ideas  from  analysing  what  happens.  
Some  examples  are  given  in  Chapter  3,  where  students  can  explore  simulations  of  
experiments   to   understand   essential   properties   of   a   frequentist   concept   of  
probability  and  perceive  that  the  variability  of  relative  frequencies  becomes  smaller  
in   longer   series   of   trials.   Working   with   such   environments   may   help   students   to  
realise   and   change   their   probability   misconceptions   that   would   persist   within   a  
formal   approach   to   probability   (see   Jones,   2005;;   Jones,   Langrall,   &   Mooney,  
2007).    
  Furthermore,   technology   offers   the   opportunity   to   students   to   learn   about  
modelling  since  it  enables  students  to  build  their  own  models  to  describe  data  or  to  
explore   probability   problems.   For   example,   a   table-­oriented   method   allows  
calculating  the  posterior  from  prior  probabilities  and  given  data,  which  is  a  method  
vital  for  a  Bayesian  decision-­oriented  approach  towards  inference  (see  Chapter  5).  
Above   all,   technological   support   may   enhance   students’   understanding   of   the  
complementary   role   of   probability   and   statistics;;   for   this   reason,   it   is   particularly  
useful  in  the  introduction  of  the  frequentist  view  of  probability  (Chapter  3)  and  in  
obtaining   empirical   sampling   distributions   with   simulation   or   re-­randomisation   to  
explore  inference  from  data  to  populations  (Chapter  5).  
  The   use   of   technology,   however,   may   also   have   drawbacks.   It   may   hide   the  
mathematical   difficulties   of   a   concept   using   artefacts   and   change   hands-­on  
activities  on  physical  objects  (e.g.,  spinners)  into  virtual  activities  (“spinners”  run  
by   software).   For   example,   drawing   a   scatter   plot   requires   quite   a   few   steps  
including  attributing  the  variables  to  the  coordinates,  deciding  the  scale  of  the  axes,  
and   plotting   the   points.   These   steps   build   an   operative   understanding   of   the   final  
plot.   If   technology   automatically   supplies   these   operations,   students   may   use   the  
default  options  (e.g.,  an  inadequate  scale)  in  an  uncritical  way.  
  Moreover,   technology   may   bias   educational   efforts.   On   the   one   hand,   teachers  
may  use  technology  to  overemphasise  computation  neglecting  the  understanding  of  
the  statistical  concepts.  On  the  other  hand,  the  learning  goals  may  be  reduced  to  the  
learning   of   simulation   techniques   not   paying   attention   to   the   concepts   or   the  
reasoning   behind   the   method   applied.   Another   drawback   of   the   overuse   of  
simulation   is   the   reduction   of   probability   to   a   mere   frequentist   concept.   To   avoid  
such  problems,  we  put  special  emphasis  on  analysing  the  learning  goals  underlying  
the   activities   proposed   for   the   work   with   the   students   in   the   classroom   and,  
throughout  the  book,  we  use  technology  as  a  complement  to  rather  than  a  substitute  
for  statistical  reasoning.  

11  
CHAPTER  1  

1.5.  ADAPTING  THE  LEVELS  OF  FORMALISATION  TO    


THE  DIVERSITY  OF  STUDENTS  

Various   factors   suggest   the   need   to   be   able   to   teach   the   same   topic   at   different  
levels  of  formalisation.  Among  them,  we  list  the  diversity  of  curricular  guidelines  
around   the   world,   the   different   educational   requirements   of   similar   high-­school  
grades   (e.g.,   technical   versus   social-­science   strands),   as   well   as   the   differing  
abilities  and  competencies  of  the  students.  As  the  book  is  intended  to  serve  a  broad  
international   audience,   we   have   tried   to   implement   the   principle   of   adapting   to   a  
diversity  of  students  throughout  the  text.    
  Let   us   consider,   for   example,   statistical   inference,   a   quite   sophisticated   topic,  
where   starting   with   an   informal   approach   has   been   recommended   to   reduce   the  
technicalities   when   introducing   the   topic   (e.g.,   Makar,   Bakker,   &   Ben-­Zvi,   2011;;  
Rossman,  2008).  This  informal  approach  could  be  used  as  an  introduction  for  the  
majority  of  students.  However,  in  some  countries  (e.g.,  in  New  Zealand,  Germany,  
or   Spain),   a   more   formal   approach   to   inference   (including   an   exposition   of  
confidence   intervals)   is   required   in   the   last   grade   of   high   school   for   particular  
strands.    
  We   therefore   start   the   exposition   of   basic   inference   methods   like   tests   of  
significance   and   confidence   intervals   in   Chapter   5   with   an   informal   approach  
where  the  sampling  distribution  is  estimated  via  simulation.  We  then  later  suggest  
that  students  with  experience  in  probability  rules  or  the  binomial  distribution  may  
use   this   previous   knowledge   to   compute   the   exact   sampling   distribution.   We   add  
further   activities   related   to   statistical   tests   as   decision   making   or   use   of   Bayes’  
theorem  to  update  prior  information  about  a  parameter  for  those  students  with  more  
advanced  experience  in  probability.  
  Other  resources  considering  the  same  curricular  content  at  different  standards  of  
formalisation   can  be  found  in  the  GAISE  guidelines  (Franklin  et  al.,  2007).  They  
provide   examples   of   how   different   levels   of   presentation   require   and   reflect   an  
increasing  sophistication  in  understanding  and  applying  stochastic  concepts.    

1.6.  STATISTICAL  AND  PROBABILISTIC  LITERACY  

Statistics  is  embedded  in  a  methodology  that  serves  to  generate  evidence  from  data.  
Statistical  knowledge  is  vital  in  research  and  in  public  discussion  where  it  is  used  
to  empower  arguments  pro  or  contra  some  issue.  Without  statistical  knowledge  it  is  
difficult  to  discern  misuse  from  proper  use  of  data.  Statistical  knowledge  involves  
thinking   in   models,   being   able   to   apply   proper   models   in   specific   situations,  
considering   the   impact   of   assumptions,   deriving   and   checking   the   results,   and  
interpreting  them  in  the  context.  

1.6.1.   Statistical  Literacy  


The  relevance  of  statistical  reasoning  and  knowledge  to  functioning  effectively  in  
the  information  society  led  to  the  introduction  of  the  term  statistical  literacy:    

12  
EDUCATIONAL  PRINCIPLES  FOR  STATISTICS  AND  PROBABILITY  

The  ability  to  understand  and  critically  evaluate  statistical  results  that  permeate  daily  life,  
coupled  with  the  ability  to  appreciate  the  contributions  that  statistical  thinking  can  make  
in  public  and  private,  professional  and  personal  decisions.  (Wallman,  1993,  p.  1)  

  Statistical   literacy   is   emphasised   in   the   GAISE   report   (Franklin   et   al.,   2007)  


produced  in  collaboration  between  the  American  Statistical  Association  (ASA)  and  
the  National  Council  of  Teachers  of  Mathematics  (NCTM,  2000).    
  Literacy   is   defined   as   the   ability   to   find,   read,   interpret,   analyse,   and   evaluate  
written  information,  and  to  detect  possible  errors  or  biases  within  this  information.  
Therefore,   to   be   statistically   literate,   people   need   a   basic   understanding   of  
statistics.   This   includes   knowing   what   statistical   terms   and   symbols   mean;;   being  
able  to  read  statistical  graphs  and  other  representations  of  data;;  understanding  the  
basic   logic   of   statistics;;   and   understanding   and   critically   evaluating   statistical  
results  that  appear  in  everyday  life.  Statistical  literacy  should  also  enable  people  to  
question   the   thinking   associated   with   a   specific   method,   to   understand   certain  
methods  and  their  limitations,  or  to  ask  crucial  questions  to  experts  and  understand  
their  answers.  For  example,  to  measure  the  success  of  some  treatment  as  compared  
to   alternative   treatments,   a   doctor   can   use   a   variety   of   criteria   such   as   life  
expectancy,  5-­year  survival  rates,  or  the  whole  survival  curve.    
  Several  researchers  have  developed  their  own  specific  models  of  competencies  
to   describe   statistical   and   probability   literacy.   We   describe   some   of   these   models  
subsequently.    

1.6.2.   Statistical  Literacy  Components  


Watson  (1997)  described  “statistical  literacy”  as  a  set  of  competencies  that  adults  
need   to   manage   “life”   in   the   information   society,   which   include   literacy,  
mathematical  and  statistical  skills,  as  well  as  knowledge  of  context  and  motivation.  
She   considered   three   levels   of   increasing   complexity:   a)   a   basic   understanding   of  
statistical   terms   and   language;;   b)   a   more   complete   understanding   of   statistical  
language  and  the  related  concepts  in  the  contexts  where  they  are  encountered;;  and  
c)   a   critical   attitude   and   thinking   ability   to   apply   statistical   ideas   to   analyse   or  
debate  statistical  claims  (see  also  Watson  &  Callingham,  2003).    
  Another  widely  accepted  description  of  statistical  literacy  was  proposed  by  Gal  
(2002)  who  identifies  two  interrelated  components  of  the  knowledge  that  an  adult  
needs  to  become  a  competent  “data  consumer”:  a)  the  person’s  ability  to  interpret  
and  critically  evaluate  statistical  information,  data-­related  arguments,  or  stochastic  
phenomena   in   diverse   contexts;;   b)   the   related   ability   to   discuss   or   communicate  
opinions   on   such   information.   These   capabilities   are   based   on   statistical   and  
mathematical   knowledge,   literacy   skills,   knowledge   of   the   context,   as   well   as  
specific  attitudes  such  as  a  critical  stance.  Gal  (2002,  p.  47)  suggests  that  the  goal  
of  statistical  literate  students  is  more  easily  reached  when  less  emphasis  is  placed  
on  the  computational  component  of  statistics  teaching:    

Some  schools  […]  teach  statistics  […]  as  part  of  mathematics,  […]  yet  not  in  a  way  that  
necessarily  emphasises  the  development  of  statistical  literacy.    

13  
CHAPTER  1  

  Gal’s   (2002)   model   includes   both   mathematical   and   statistical   knowledge.  


Reading  tables  and  graphs  already  requires  mathematical  competencies  as  well  as  
taking  a  critical  stance  in  regard  to  statistical  arguments.  Mathematical  thinking  is  
also  included  in  modelling  and  understanding  the  definition  of  particular  variables  
when   collecting   and   analysing   data.   For   example,   unemployment   is   defined  
differently  in  various  countries,  poverty  is  defined  relative  to  a  population,  and  the  
proportion   of   poor   does   not   change   if   all   people   get   richer   by   a   fixed   amount   of  
money.    
  When   statisticians   communicate   results   to   the   target   audience,   they   do   not  
always  explain  all  of  their  assumptions;;  hence,  the  data  may  appear  as  if  they  were  
facts.   Nevertheless   the   validity   of   results   depends   on   these   assumptions.  
Maintaining  a  balance  between  simplifying  the  communication  of  statistical  results  
and   preserving   authentic   statistical   information   is   difficult:   for   in   simplifying   the  
information   to   reach   more   people,   the   statistician   might   disengage   those   who   are  
really  interested  in  obtaining  more  complete  information.  
  Gal  (2005)  expanded  these  ideas  to  probability  literacy.  This  extension  includes  
the   capability   to   interpret   and   critically   evaluate   probabilistic   information   and  
random  phenomena  in  diverse  contexts.  Essential  for  such  literacy  are  the  abilities  
to   understand   the   meaning   and   language   of   basic   probability   concepts   and   to   use  
probability   arguments   properly   in   private   or   public   discussions.   Gal   includes  
dispositional   elements   in   his   description   of   probability   literacy,   e.g.,   appropriate  
beliefs  and  attitudes,  and  control  of  personal  feelings  such  as  risk  aversion  without  
defensible  reasons.    

1.6.3.   Actions  and  Resources  Directed  to  Increase  Statistical  Literacy  


Vere-­Jones   (1995)   documented   the   early   initiatives   to   promote   statistical   literacy.  
The   Education   Committee   was   founded   in   1948   as   joint   initiative   of   the  
International   Statistical   Institute   (ISI)   and   the   UN.   The   Taskforce   on   Teaching  
Statistics  at  School  Level,  established  by  the  ISI  in  1971,  organised  a  Round  Table,  
which   analysed   the   international   status   of   teaching   statistics   at   school   level  
(Barnett,  1982).    
  To  improve  statistical  education  around  the  world,  the  ISI  launched  a  series  of  
conferences   in   1982   under   the   title,   International   Conference   on   Teaching  
Statistics   (ICOTS;;   for   the   latest,   see   Makar,   Sousa,   &   Gould,   2014).   Another  
milestone   was   the   establishment   of   the   International   Association   for   Statistical  
Education   (IASE)   in   1991   as   a   separate   section   of   the   ISI   (iase-­web.org).   The  
founding   of   specific   journals11   inspired   and   facilitated   the   exchange   of   ideas   and  
consolidated  the  young  research  community.    
                                                                                                                         
11
    The   main   journals   are   Statistics   Education   Research   Journal   (iase-­web.org/  
Publications.php?p=SERJ),   Journal   of   Statistics   Education   (www.amstat.org/publications/jse/),  
Teaching   Statistics   (www.teachingstatistics.co.uk),   Statistique   et   Enseignement   (publications-­
sfds.math.cnrs.fr/index.php/StatEns),   Technology   Innovations   in   Statistics   Education  
(escholarship.org/uc/uclastat_cts_tise)  and  Stochastik  in  der  Schule  (stochastik-­in-­der-­schule.de/).    

14  
EDUCATIONAL  PRINCIPLES  FOR  STATISTICS  AND  PROBABILITY  

  A  major  initiative  towards  statistical  literacy  for  everybody  is  the  International  
Statistical   Literacy   Project   promoted   by   the   ISI   and   the   IASE.   The   activities  
include   competitions   for   students,   the   World   Statistics   Day,   the   best   cooperative  
project   competition,   a   newsletter,   and   an   electronic   repository   of   teaching  
resources.  

1.7.  STATISTICAL  AND  PROBABILISTIC  THINKING  

In  the  previous  sections,  we  discussed  the  specificity  of  statistical  and  probabilistic  
thinking.  Stochastic  concepts  are  derived  by  arguments  with  a  strong  mathematical  
component.   Nowhere   else   in   mathematics,   has   the   overlap   between   the   scientific  
and   philosophical   ideas   been   as   strong,   and   these   ideas   were   only   separated  
relatively   recently.   Moreover,   stochastics   has   shaped   modern   research  
methodology  and  is  now  an  integral  part  of  empirical  research  as  shown  in  Popper  
(1959).  
  A   source   of   difficulty   is   the   mixture   of   theoretical   ideas   and   personal  
conceptions  when  people  apply  stochastic  reasoning.  One  example  of  this   is  the  
complex   relationship   between   randomness   and   causality.   In   a   causal   situation,  
the   outcome   is   completely   determined   by   the   conditions   of   an   experiment;;  
however,  in  a  stochastic  situation  there  is  no  explicit  way  to  predict  the  outcome  
with  certainty,  even  though  the  experiment  is  thought  to  be  repeatable  under  the  
same  conditions  as  in  physics.  The  assumption  is  that  the  repetitions  of  a  random  
experiment   are   “independent”.   Within   probability   theory,   independence   is  
reduced  to  the  multiplicative  rule.  When  this  rule  has  to  be  applied  to  a  situation  
with  data  in  a  real  context,  one  has  to  motivate  why  the  specific  experiment  can  
be  modelled  as  an  independent  stochastic  experiment.  However,  many  textbooks  
refer   to   independence   as   a   lack   of   causal   influence,   which   only   adds   to   the  
confusion.  
  When   teaching   these   concepts,   often   only   the   mathematical   definitions   are  
developed  while  the  intuitive  side  is  neglected.  Probabilistic  modelling  (described  
in  Chapter  3),  however,  has  to  rely  on  some  basic  assumptions  that  all  methods  of  
statistical   inference   have   in   common;;   consequently,   without   a   connection   to  
probability,   statistics   can   be   deprived   not   only   of   its   potential   in   applications   but  
also  of  its  sense-­making.  
  Consistent  with  the  ideas  above,  many  authors  have  pointed  to  the  unique  nature  
of  statistics  and  probability  (e.g.,  Moore  &  Cobb,  2000;;  Scheaffer,  2006)  whence  it  
is  wise  to  use  a  modelling  approach  for  teaching.  Modelling  includes  working  with  
real  data,  searching  for  suitable  variables  to  answer  questions  from  a  context  (see  
Franklin  et  al.,  2007),  measuring  objects  on  the  scale  of  the  variables,  controlling  
for   variation   in   data   by   the   selection   of   samples,   and   reducing   the   effects   of  
confounding  factors  by  intelligent  design  of  data  production.    

15  
CHAPTER  1  

1.7.1.   Statistical  Thinking  


Thinking   is   something   human   beings   do   all   the   time   and   it   is   influenced   by   their  
experiences   and   the   theoretical   frameworks   they   have   acquired   in   their   life.  
Mathematical  thinking  is  thinking  influenced  by  a  mathematical  background.  It  can  
be   formal   if   the   mathematical   notions   and   the   relationships   are   precisely   used;;  
however,  according  to  Fischbein  (1987)  mathematicians  also  have  an  intuitive,  yet  
mathematical   way   of   thinking.   In   that   case   their   thinking   is   guided   by   secondary  
intuitions   that   emerge   from   learning   the   mathematical   concepts,   replacing   their  
primary   intuitions   that   exist   prior   to   formal   education.   Such   secondary   intuitions  
allow   one   to   find   short-­cuts   for   a   solution   of   a   mathematical   problem.   Thinking  
mathematically   comprises   finding   a   suitable   model   to   adequately   represent   the  
situation.   Of   course,   in   the   modelling   process   the   constituents   are   wider   and   also  
include   –   beyond   mathematics   –   knowledge   of   the   context,   as   well   as   criteria   for  
assessing  how  well  models  match  a  situation.  
  Stochastic   thinking   has   been   described   by   different   authors,   starting   from  
Heitele  (1975)  who  described  a  list  of  fundamental  ideas  related  to  understanding  
probability.   One   influential   example   of   a   modelling   framework   for   statistical  
thinking  was  developed  by  Wild  and  Pfannkuch  (1999)  who  described  the  complex  
thought   processes   involved   in   solving   real-­world   problems   using   statistics.   They  
used   the   modelling   cycle   described   by   the   modelling   group   within   mathematics  
education   (e.g.,   in   Blum,   1985,   or   in   Blum,   Galbraith,   Henn,   &   Niss,   2007)   and  
filled  its  components  with  statistical  ideas.  They  mainly  focused  on  the  process  of  
empirical   research   with   little   reference   to   probability   although   it   is   possible   to  
interpret   their   description   to   include   probabilistic   models.   They   developed   a  
complex  framework  including  the  following  components  that  we  briefly  summarise  
under  the  header  statistical  investigation  cycle.  

1.7.2.   The  Statistical  Investigation  Cycle    


Solving   statistical   problems   involves   a   complete   research   cycle   of   Problem,  
Planning,   Data,   Analysis,   and   Conclusion   (PPDAC).   A   cycle   (Figure   1.1)   begins  
when  a  person  conceptualises  a  statistical  problem  embedded  in  a  larger  real-­world  
problem.  Various  examples  are  given  throughout  the  book  such  as  investigating  the  
differences   between   the   physical   measures   of   boys   and   girls   in   a   classroom  
(Chapter  2),  deciding  about  the  convenience  of  an  insurance  contract  (Chapter  3),  
predicting  life  expectancy  from  other  variables  (Chapter  4),  or  fixing  an  acceptable  
proportion   of   defective   items   in   quality   control   (Chapter   5).   Most   of   these  
problems  emerge  from  a  practical  need,  where  statistics  is  used  to  find  a  reasonable  
answer.  

Problem.   A  solution  to  the  real  problem  requires  that  the  student  understands  the  
context   and   realises   what   is   the   relevant   question   to   be   answered.   Often   this  
original   question   is   too   wide   and   needs   to   be   made   more   precise;;   for   example,  
restricting   the   possible   decisions   in   the   insurance   contract.   The   first   step   in   the  

16  
EDUCATIONAL  PRINCIPLES  FOR  STATISTICS  AND  PROBABILITY  

modelling   cycle   is   refining   the   problem   in   such   a   way   that   it   can   be   posed   in  
statistical  terms;;  it  is  necessary  to  get  familiar  with  the  context  and  fix  the  specific  
goals  that  should  be  met  in  solving  the  problem.    

Planning.   Once   the   problem   is   manageable,   students   have   to   think   ahead   about  
how   to   develop   a   strategy   to   get   a   solution.   In   case   the   solution   to   the   problem  
depends   on   some   quantity   that   varies   with   the   members   of   a   specific   population  
(shoe  size,  in  Chapter  2),  it  is  necessary  to  plan  how  to  measure  this  quantity  across  
the   population   members   (boys   and   girls).   Then   there   is   the   decision   about   which  
data   should   be   collected   (number   of   students,   specific   groups   in   the   sample)   and  
which   strategy   is   followed   to   sample   the   population   (systematic   or   random  
sample).   Finally,   a   plan   of   the   analysis   is   required   (in   this   example,   a   descriptive  
analysis  is  appropriate).  
 
– Understanding  
 
Problem – Refining  

   
– Interpretation   – Measurement  
– Solutions   – Data  collection  
– New  problems   Conclusion Planning – Sampling  
– Communication   – Analysis  
   
   
– Exploratory  analysis  
– Collection  
– Generating  hypotheses   Analysis Data
– Cleaning  
– Testing  hypotheses     – Organisation  
– Estimation/prediction  

Figure  1.1.  The  PPDAC  cycle  (adapted  from  Wild  &  Pfannkuch,  1999,  p.  226)  

Data.   The   next   step   is   collecting   the   data.   Sometimes   the   students   collect  
accessible   data   (e.g.,   from   the   Internet)   or   use   data   from   previous   questionnaires.  
Other  times,  the  students  design  questionnaires  or  experiments  to  collect  their  own  
data.  It  has  to  be  mentioned  that  their  statistical  understanding  increases  when  they  
are   involved   in   the   design   of   a   data   collection   instrument   or   experiment.   An  
example   where   the   students   are   involved   in   organising   an   experiment   to   collect  
data  to  test  a  hypothesis  is  given  in  Chapter  5.  Once  the  data  are  collected,  they  are  
placed   in   a   spreadsheet   or   other   software   that   facilitates   computations   and  
visualisations.   This   sometimes   requires   reorganising   the   data,   for   example,   in  
selecting   the   variables   of   interest   and   deciding   the   units   to   use   for   the   analysis.  
Data  should  also  be  inspected  for  errors  in  measurement  or  recording.  

17  
CHAPTER  1  

Analysis.   After   the   data   are   collected   and   cleaned,   the   analysis   can   start.   If   the  
students  have  no  idea  about  what  can  be  found  in  the  data,  they  might  carry  out  an  
exploratory   data   analysis   (EDA).   In   this   approach   (see   Chapters   2   and   4),   the  
students   summarise   the   characteristics   of   the   data   without   testing   pre-­set  
hypotheses.  Tukey  (1977)  was  the  main  promoter  of  this  approach,  where  multiple  
data  representations  and  visualisations  are  used  to   discover  hidden  patterns  in  the  
data   and   to   formulate   new   hypotheses.   Other   times   the   students   perform   an  
inferential   analysis   (see   Chapter   5).   In   this   case   the   interest   is   generalising   a  
conclusion  from  the  data  at  hand  to  a  wider  population  from  which  the  data  are  a  
random  sample.  Inference  is  used  to  make  estimations  or  predictions  about  a  larger  
population   that   the   sample   represents   or   to   test   hypotheses   set   prior   to   collecting  
the  data.  

Conclusion.   The   final   step   in   the   modelling   cycle   is   the   interpretation   of   the  
results   from   the   analysis   and   relating   this   interpretation   to   the   context   in   such   a  
way  that  produces  some  answers  to  the  original  problem.  For  example,  in  the  tea-­
tasting   experiment   to   test   a   hypothesis   in   Chapter   5,   we   obtain   data   that   are   very  
unlikely  if  the  hypothesis  is  true.  We  interpret  the  finding  as  evidence  against  the  
initial   hypothesis   and   therefore   reject   it.   In   Chapter   2,   we   decide   to   repeat   the  
analysis  by  discarding  some  atypical  values.  The  modelling  cycle  can  be  repeated  
several  times  until  a  reasonable  conclusion  is  reached.  

1.7.3.   Fundamental  Types  of  Statistical  Thinking  


Carrying  out  a  statistical  research  process  requires  that  the  person  continually  poses  
and   solves   new   questions.   Wild   and   Pfannkuch   (1999)   defined   specific   types   of  
thinking  processes  that  constantly  appear  during  the  PPDAC  cycle.  Pfannkuch  and  
Wild  (2004)  defined  these  fundamental  types  of  thinking  as  follows:    

1.   Appreciation   for   the   need   for   data.   As   suggested   in   Section   1.2,   data   are  
essential   in   statistics;;   very   few   statistical   investigations   can   be   completed  
without   properly   collected   and   analysed   data.   Even   if   people   have   their   own  
experience  with  the  same  type  of  situation  they  should  not  base  a  solution  solely  
on  their  personal  experience.  Reliable  data  are  essential  to  provide  information  
for   a   solution   and   to   reach   an   adequate   decision   or   judgement   about   the  
situation.  An  important  part  of  statistical  thinking  is  being  able  to  recognise  the  
points  in  the  process  where  new  data  are  needed.  

2.   Transnumeration.  By  summarising  and  visualising  data,  patterns  and  deviations  


from  these  patterns  (hidden  in  the  raw  data)  might  be  discovered.  Changing  the  
data  representation  may  enhance  the  understanding  of  the  situation.  The  process  
of  discovering  ideas,  tendencies,  or  structure  from  the  data  by  changing  the  data  
representation  is  called  transnumeration  by  Pfannkuch  and  Wild  (2004).      
For   example,   by   measuring,   some   characteristics   of   the   situation   are   captured  
(e.g.,   different   heights   in   a   sample   of   students,   different   hair   colours,   etc.).  

18  
EDUCATIONAL  PRINCIPLES  FOR  STATISTICS  AND  PROBABILITY  

Visualising  the  data  in  a  bar  graph  (if  applicable)  shows  the  mode  (or  modes  if  
there  is  no  unique  peak)  and  the  range  of  the  variables.  Changing  to  a  box  plot,  
the   median,   the   quartiles,   and   the   extreme   values   become   visible   (see   Chapter  
2).  The  change  of  representations  might  reveal  new  relevant  information  in  the  
data.    

3.   Perception  of  and  attention  to  sources  of  variation.  Variation  occurs  in  all  areas  
of  mathematics;;  however,  random  variation  is  specific  to  statistics.  A  particular  
type   of   statistical   thinking   is   to   differentiate   statistical   and   non-­statistical  
(deterministic)  variation.  It  is  also  important  to  recognise  the  various  sources  of  
variation:  natural  variation  in  the  population,  error  in  measurement,  or  variation  
in  sampling  of  data.      
A   goal   of   statistics   is   to   separate   irreducible   and   reducible   variation.   Even  
natural   variation   in   the   population   can   be   reduced;;   for   example,   in   Chapter   2,  
the  analysis  of  arm  span  should  be  separated  between  boys  and  girls.  Statistics  
offers   methods   to   control   variation   when   the   source   of   variability   is   known.  
Variation   is   also   inherent   in   the   conclusions;;   for   example,   a   p-­value   or   a  
confidence   level   indicate   the   quality   of   the   used   statistical   argument   based   on  
samples  from  the  population  (see  Chapter  5).  

4.   Integration  of  statistics  and  context.  Contextual  knowledge  is  vital  in  all  steps  of  
the  modelling  cycle.  To  highlight  its  relevance,  Wild  and  Pfannkuch  include  the  
integration  of  statistics  and  context  as  a  specific  type  of  statistical  thinking.  The  
statistical  model  must  be  selected  and  exploited  in  such  a  way  that  the  essential  
elements  of  the  real  situation  are  captured.  At  the  same  time  it  is  necessary  to  be  
conscious   that   any   model   is   different   from   reality;;   hence,   some   differences  
between  the  model  and  the  investigated  problem  situation  remain.  The  target  is  
to   generate   data   that   contain   adequate   information   needed   to   answer   the   initial  
problem;;   the   summary   report   should   be   oriented   to   synthesise,   understand   and  
generalise   the   situation,   when   possible.   Most   importantly,   integration   with   the  
context   is   essential   in   the   conclusion   phase   where   it   is   decided   whether   the  
solution  is  reasonable  and  applicable  in  the  context.  

5.   Using  appropriate  statistical  models.  As  in  other  areas  of  applied  mathematics,  
modelling   is   essential   in   statistics.   The   opportunities   for   modelling   are  
unbelievably   wide   in   statistics   (as   seen   from   the   examples   in   Tanur,   1989).  
Moreover,  statistics  has  developed  its  own  set  of  models  that  were  specifically  
developed   for   the   analysis   of   data.   There   is   a   wide   range   of   statistical   models,  
some   of   which   are   highly   sophisticated.   Examples   include   the   normal  
distribution,   regression   models,   or   statistical   tests,   which   can   be   generalised   to  
complex   situations.   However,   it   is   possible   to   use   simple   versions   of   these  
models  at  high-­school  level  (see  Chapters  3  to  5).  A  feature  of  statistical  models  
described   by   Wild   and   Pfannkuch   is   that   they   help   us   to   think   in   terms   of  
distributions  (aggregates)  instead  of  concentrating  on  individuals.  

19  
CHAPTER  1  

1.7.4.   Components  of  Probabilistic  Thinking    


Borovcnik  (2011)  described  probabilistic  thinking  by  elaborating  on  the  following  
components:  a)  the  ability  to  balance  between  psychological  and  formal  elements  
of  probability  when  using  a  personal  scale  of  probability;;  b)  the  understanding  that  
there   are   no   direct   success   criteria   in   random   situations;;12   c)   the   ability   to  
discriminate  randomness  and  causality;;  and  d)  the  understanding  of  the  difference  
between   reflecting   on   a   problem   and   making   a   decision.   Other   components   of  
probabilistic  thinking  include:  

1.   Influence   of   prior   probability   judgement.   Realising   that   many   probabilities   are  


dependent   on   other   (prior)   probabilities   and   for   this   reason   these   probabilities  
should   be   related   to   the   proper   subgroup.   For   example,   the   probability   of   a  
woman  with  a  positive  mammogram  having  breast  cancer  depends  on  the  prior  
probability  of  having  breast  cancer,  which  describes  the  risk  of  “her”  subgroup  
(Chapter  3).    

2.   Asymmetry   of   conditional   probabilities.   Understanding   that   conditional  


probabilities   establish   a   non-­symmetric   relation   between   events   is   a   key   to  
dealing   with   probabilities   and   interpreting   them   properly.   For   example,   if   the  
probability   of   a   positive   mammogram  is  high  given  that  the  woman  has  breast  
cancer  that  does  not  imply  that  the  reverse  conditional  probability  is  also  high.  It  
is   vital   for   probabilistic   thinking   to   be   able   to   relate   the   reversed   conditional  
probability  to  the  prior.  

3.   Theoretical   character   of   independence.   Applying   independence   is   often   an  


inherent   requirement   of   probabilistic   models   but   is   hard   to   check   whether   it   is  
really   appropriate.   For   example,   independence   usually   cannot   be   applied   when  
two  pieces  of  circumstantial  evidence  are  combined  at  court.  Despite  its  abstract  
meaning,   independence   is   a   main   concept   in   probability   as   frequencies   make  
sense  only  if  experiments  are  “independently”  performed  (Chapter  3).  

4.   The   problem   of   small   probabilities.   Interpreting   small   probabilities  


appropriately   is   extremely   difficult.   Small   (conditional)   probabilities   of  
observed   results   are   used   to   reject   the   hypothesis   in   inference   (see   Chapter   5)  
but   this   does   not   establish   a   contradiction   to   the   hypothesis   (in   the   sense   of  
logic).  It  is  difficult  but  essential  to  understand  the  logic  of  a  significance  test  as  
it   weakens   mathematical   proof   to   empirical   evidence.   Small   probabilities   also  
occur  in  the  evaluation  of  risks  and  are  difficult  to  handle  because  the  researcher  
usually   does   not   have   enough   data.   For   example,   Dubben   and   Beck-­Bornholdt  

                                                                                                                         
12
    The  person  should  understand  that  a  correct  strategy  does  not  always  assure  a  success.  

20  
EDUCATIONAL  PRINCIPLES  FOR  STATISTICS  AND  PROBABILITY  

 
(2010)   describe   an   epidemiological   study   where   all   331   BSE13 positive   cases  
could  be  false  positives  as  there  is  no  proper  estimate  of  BSE  prevalence.  

5.   Correlation  as  probabilistic  dependence.  Understanding  that  correlation  is  based  


on   probabilities   and   conceptualises   a   much   weaker   form   of   relationship   than  
functional  dependence.  Furthermore,  it  is  important  to  accept  that  correlation  can  
be   increased,   generated,   or   changed   by   other   variables   or   artefacts14   which   are  
often   neglected   (see   also   Chapter   4).   A   proper   interpretation   of   correlation  
marks  a  great  step  towards  thinking  probabilistically.  

1.8.  MAKING  SENSE  OF  STATISTICS  AND  PROBABILITY  

Often,  teachers  present  statistical  information  such  as  definitions  of  new  concepts  
or  examples  of  procedures  for  solving  statistical  problems,  and  then  give  exercises  
to   the   students   to   practice   what   they   learnt.   The   consequence   is   routine   learning  
where   students   apply   the   formulas   without   any   deeper   understanding   of   the  
underlying  concepts.  
  In  order  to  improve  the  situation,  statistics  educators  recommend  refocusing  the  
teaching  of  statistics  on  reasoning  and  sense  making  (e.g.,  Shaughnessy,  Chance,  &  
Kranendonk,   2009).   According   to   the   Common   Core   State   Standards   Initiative  
(CCSSI,  2010),  to  make  sense  of  the  problems  posed  to  them,  the  students  should  
first   understand   the   goals   and   constraints,   and   conjecture   a   possible   solution   path  
before  starting  to  solve  the  tasks  posed.  They  may  consider  similar  problems   that  
have   been   solved   before   or   solve   a   simpler   form   of   the   original   problem   (e.g.,  
reduce   the   sample   size).   The   teacher   may   provide   support   when   needed;;   for  
example,  suggesting  the  use  of  a  particular  type  of  graph  to  discover  patterns  and  
relationships   in   the   data.   Another   strategy   is   organising   the   students   in   groups,  
where  more  advanced  students  help  their  classmates.    
  Throughout   the   book,   we   attempt   to   make   sense   of   the   different   concepts   and  
methods   by   using   contexts   where   these   ideas   can   be   meaningful   for   the   students.  
For  example,  in  Chapter  3,  we  use  contexts  familiar  to  students  to  introduce  three  
different   views   of   probability:   a)   probability   as   a   value   to   which   the   relative  
frequency   tends   in   a   large   number   of   experiments;;   b)   probability   as   a   ratio   of  
favourable  to  possible  cases  when  the  elementary  events  are  equally  likely;;  and  c)  
probability   as   a   personal   degree   of   belief.   In   the   same   chapter,   conditional  
probability  is  linked  to  circumstantial  evidence.  In  Chapters  3  and  5,  Bayes’  rule  is  
introduced   as   a   method   to   learn   from   experience.   The   correlation   coefficient   is  
related  to  both  the  error  in  prediction  and  the  spread  of  scatter  plots  in  Chapter  4.    
 
 
                                                                                                                         
13
    Bovine  spongiform  encephalopathy  or  mad  cow  disease,  a  fatal  neurodegenerative  disease  in  cattle.  
14
    Regression   of   average   (aggregate)   values   instead   of   using   the   original   data   can   increase   the  
correlation   considerably;;   by   the   Simpson   paradox   a   positive   correlation   in   all   subgroups   may   be  
changed  to  a  negative  correlation  in  the  whole  group  investigated.  

21  
CHAPTER  1  

We   also   try   to   build   on   students’   previous   knowledge   to   develop   a   deeper  


understanding   of   statistical   concepts.   For   this   reason,   each   concept   is   introduced  
via   tasks   that   reveal   how   statistics   and   probability   concepts   may   help   to   solve   a  
given  problem.    

1.9.  STATISTICAL  INVESTIGATIONS  AND  EXPERIMENTS  

A   traditional   teaching   approach   based   on   structured   lessons   and   simple   exercises  


can  miss  the  goals  of  developing  statistical  and  probability  literacy  and  fail  to  make  
sense   from   concepts   and   procedures   for   students.   Alternatively,   experiments   and  
projects   have   been   recommended   as   substantial   working   modes   for   teaching  
statistics   and   probability   (e.g.,   McGillivray   &   Pereira-­Mendoza,   2011).   Such  
activities  may  be  linked  to  real-­world  problems  and  real  or  simulated  data.  Students  
can  work  –  alone  or  in  small  groups  –  on  their  own  projects  and  experiments.  The  
focus  may  be  put  on  practical  work  according  to  the  cycles  of  statistical  modelling  
in  empirical  research  as  suggested  by  Wild  and  Pfannkuch  (1999).    
  For   example,   in   Chapter   4,   the   context   of   life   expectancy   is   used   to   introduce  
scatter   plots,   the   correlation   coefficient,   regression   lines,   and   the   coefficient   of  
determination.   Within   such   a   project-­oriented   approach,   students   can   a)   develop  
and  reinforce  their  capabilities  to  formulate  questions;;   b)  design  and  implement  a  
data   collection   plan   for   statistical   studies   including   observational   studies,   sample  
surveys,  and  comparative  experiments;;  and  c)  justify  conclusions  that  are  based  on  
data   (see,   e.g.,   Australian   Curriculum,   Assessment   and   Reporting   Authority  
[ACARA],  2013;;  NCTM,  2000).    
  In  Chapter  5,  an  experiment  about  testing  the  possibility  of  merely  guessing  the  
order  tea  and  milk  are  added  in  a  cup  of  tea  serves  as  a  framework  to  introduce  the  
relevant  inferential  concepts  included  in  the  high-­school  curricula.  Other  examples  
of   projects   and   investigations   are   included   throughout   this   book.   We   attempt   to  
show   how   projects   and   investigations   are   a   catalyst   for   student   engagement,   for  
learning  to  solve  problems  in  a  given  context,  for  developing  statistical  reasoning  
competencies,   and   for   synthesising   learning.   Above   all,   projects   facilitate   that  
students  can  make  sense  of  statistical  work.  
  Each  of  the  content  chapters  (Chapters  2  to  5)  is  devoted  to  the  elaboration  of  a  
set  of  related  fundamental  ideas  that  are  grounded  in  the  educational  principles  of  
the   first   chapter.   We   introduce   concepts   and   procedures   via   investigations   or  
experiments   to   assist   students,   classroom   teachers,   and   teacher   educators   to  
understand  the  need  for  the  stochastic  ideas  that  should  be  learned.    

1.10.  FINAL  THOUGHTS  

In   this   chapter,   we   have   described   the   educational   assumptions   inherent   in   the  


exposition   of   the   ideas   throughout   the   book.   Teachers   need   specific   support   to  
facilitate   students   acquire   the   related   statistical   and   probabilistic   ideas,   to  
appreciate   and   acknowledge   the   complementary   features   of   mathematical   and  

22  
EDUCATIONAL  PRINCIPLES  FOR  STATISTICS  AND  PROBABILITY  

statistical  knowledge  (and  reasoning),  to  learn  to  make  the  best  use  of  technology,  
and  to  orientate  their  teaching  to  students  of  differing  abilities.    
  The   aim   of   this   book   is   to   sustain   teacher   educators   and   teachers   as   well   as   to  
increase   their   interest,   competencies,   and   knowledge   in   stochastic   education.   We  
hope  to  see  that  education  researchers  are  encouraged  to  explore  innovative  ways  
and  tools  for  educating  teachers  and  students  in  statistics  and  probability  using  the  
ideas  suggested  in  this  book.  
   

23  

You might also like